AI search is a big deal. AI Overviews matter. AI Mode matters. Generative Engine Optimization, Answer Engine Optimization, citations, query fan-out, and whatever acronym we invent next are worth understanding.
But there is something deeply funny about watching the SEO industry debate the future of retrieval while somebody in the next tab is still keyword stuffing an H2, cannibalizing their own pages, breaking internal links, or changing a title tag every time Semrush sneezes.
That is where Evergreen Lantern SEO came from.
Sure, the memes are jokes. But the search engine sins underneath them are not.
SEO fundamentals still matter in the age of AI search because generative search systems have not replaced crawling, indexing, relevance, useful content, internal linking, technical clarity, or good editorial judgment. Google explicitly says its generative AI search features are rooted in its core Search ranking and quality systems.
AI changes how information may be retrieved, synthesized, cited, and presented, but it does not rescue bad architecture, thin content, misunderstood intent, manipulative links, or sloppy implementation. Learn the new systems, absolutely. Just do not forget how the web works.
Key Takeaways
- AI search sits on top of SEO fundamentals. Crawlability, indexability, relevance, useful content, and clear site structure still matter.
- Most of these sins are not beginner mistakes. They are things experienced practitioners already know, which somehow makes them funnier.
- Use AI-era tactics after the foundations work. An
llms.txtfile will not save a site whose important pages cannot be crawled.
1. Please Stop Keyword Stuffing Like Semantic Search Never Happened
[Insert Evergreen Lantern Meme #1]
Evergreen Lantern says:
Please stop forcing the exact-match keyword into every H2 because your content optimization score turned orange.
Keyword stuffing has survived an impressive number of SEO eras.
We went from primitive search engines to machine learning, semantic systems, neural matching, natural-language understanding, and generative search, yet somebody somewhere is still trying to squeeze the primary keyword into a paragraph for the seventh time because a plugin turned a circle yellow.
Using the language your audience searches for is still good SEO. Repeating it unnaturally is not.
Google continues to classify keyword stuffing as spam when keywords or numbers are inserted unnaturally in an attempt to manipulate rankings.
The useful distinction is relevance versus repetition.
Use important terminology where it helps the reader and clearly establishes the page topic. Then write normally. Synonyms, related concepts, entities, examples, and natural language do a much better job of demonstrating subject understanding than repeating one string until the copy sounds broken.
Do this instead: Identify the main query and intent, cover the subject comprehensively, and let the language vary naturally.
Authority source: Google Search Central: Spam Policies for Google Web Search
2. Please Stop Changing the Title Tag Every Time the Rank Tracker Twitches
[Insert Evergreen Lantern Meme #2]
Evergreen Lantern says:
Please stop changing the title tag every three days because Semrush moved you from 7.2 to 7.8.
This one deserves to be framed.
Optimization requires iteration. Compulsive optimization is something else.
A ranking tracker moves. Somebody changes the title. Two days later the number moves again. Somebody changes it back. Then the H1 gets involved. Eventually nobody knows which change caused what because the page has been subjected to an uncontrolled SEO chemistry experiment.
Google does not simply read your new title tag and instantly lock it into the SERP. Its title links are generated automatically from several signals, and Google says changes may require recrawling and reprocessing that can take anywhere from days to weeks.
That makes rapid-fire tweaking especially bad experimental design.
If you are testing a title, give the change enough time and enough impressions to produce usable information.
Do this instead: Make deliberate title changes based on intent, relevance, CTR opportunity, or a genuine content change. Record the change, establish your comparison period, and stop touching it long enough to learn something.
Authority source: Google Search Central: Influencing Your Title Links in Search Results
3. Please Stop Making a New URL Every Time the Keyword Changes Slightly
[Insert Evergreen Lantern Meme #3]
Evergreen Lantern says:
Please stop making “best CRM for dentists,” “best dental CRM,” and “CRM software for dentists” three different fucking pages.
There was a time when exact-match keyword targeting encouraged some spectacular website architecture.
One keyword. One URL.
Plural version? New URL.
Different adjective? New URL.
Different preposition? Apparently we need another landing page.
Modern search systems are considerably better at understanding meaning than that.
Google’s current generative-search guidance specifically warns against creating separate content for every possible search or query fan-out variation. It also says its systems can understand relevance even when the page does not exactly match the wording of the query.
This is not permission to ignore keyword research. It is a reminder of what keyword research is actually for.
Keywords help us understand demand. Intent helps us decide what deserves a page.
If three queries represent the same problem, same reader, same expected answer, and same useful destination, they probably do not need three URLs.
Do this instead: Cluster queries by actual intent before assigning URLs. Create another page when the reader needs meaningfully different information, not because Ahrefs found another grammatical variation.
Authority source: Google Search Central: Optimizing Your Website for Generative AI Features
4. Please Stop Calling 37 Versions of the Same Article “Topical Authority”
[Insert Evergreen Lantern Meme #4]
Evergreen Lantern says:
Please stop calling it topical authority. You published the same article 37 times with different title tags.
Topical depth is useful.
Topical duplication is not depth.
A good content cluster expands the reader’s understanding. Each page earns its existence by handling a distinct question, task, comparison, use case, or level of detail.
A bad cluster takes one idea and reproduces it until the sitemap looks authoritative from a distance.
This problem is likely to get worse because AI makes content production cheap. A team can now create 50 technically competent pages faster than it once created five.
That does not mean the site has gained 10 times more expertise.
Google’s current AI-search guidance is unusually direct here: a higher quantity of pages does not make a site higher quality or more relevant. It also recommends reducing duplicate content and focusing on unique, useful material.
Do this instead: Before adding another article to a cluster, ask what new job that page performs. If the answer is basically “rank for another variation,” consider improving or consolidating what already exists.
Authority source: Google Search Central: Optimizing Your Website for Generative AI Features
5. Please Stop Publishing AI Content That Has Nothing New to Contribute
[Insert Evergreen Lantern Meme #5]
Evergreen Lantern says:
Please stop publishing 40 AI articles a week when none of them contains a single thing worth citing.
The problem is not that AI touched the content.
I use AI. Plenty of good writers, strategists, researchers, and marketers use AI.
The problem is commodity content at industrial scale.
If an article merely reorganizes information that already exists across 500 competing pages, producing it faster does not make it more valuable. It only gets us to mediocrity more efficiently.
Google’s spam policy describes scaled content abuse as producing large amounts of unoriginal material primarily to manipulate rankings instead of helping users, regardless of how that content was created. Its separate guidance on generative AI says AI can be useful for research and structure, but mass-generating pages without adding value may violate that policy.
Google’s newer generative-search guidance goes even further by emphasizing non-commodity content, including firsthand experience, unique viewpoints, and material that does more than restate what is already available.
That is probably even more important when AI systems themselves can synthesize common knowledge.
Do this instead: Use AI to accelerate the parts worth accelerating. Then add something the model could not have pulled from everybody else: experience, original data, interviews, examples, analysis, process knowledge, testing, informed opinion, or proprietary context.
Authority sources:
6. Please Stop Treating an Optimization Score Like Editorial Judgment
[Insert Evergreen Lantern Meme #6]
Evergreen Lantern says:
Please stop telling me the content is helpful because Surfer gave it a 93.
SEO tools are useful.
I use SEO tools.
They help us identify terms, gaps, competitor patterns, internal-link opportunities, technical problems, and questions worth investigating.
But a score is still a model built by somebody other than Google attempting to approximate something useful.
It is not a certificate from the ranking gods.
A page can achieve an impressive optimization score while remaining dull, redundant, inaccurate, poorly structured, or completely wrong for the user.
Google now explicitly warns site owners to be cautious about third-party products that imply access to internal Google ranking or AI metrics. No third-party tool has access to those systems.
That does not make the tools useless. It puts them in their proper place.
Tools assist judgment. They do not replace it.
Do this instead: Use scoring tools diagnostically. Ask what a recommendation reveals, decide whether it improves the page, and ignore it when satisfying the tool would make the content worse.
Authority source: Google Search Central: Optimizing Your Website for Generative AI Features
7. Please Answer the Damn Question Before the Eighth H2
[Insert Evergreen Lantern Meme #7]
Evergreen Lantern says:
Please answer the question somewhere above the eighth H2. Nobody asked for 900 words about today’s rapidly evolving digital landscape.
SEO writers developed some strange habits during the long-form-content era.
One of them was treating length as proof of completeness.
That is how someone searching for a straightforward answer ends up reading the history of plumbing before learning how to unclog a sink.
Long content is not inherently bad. Some questions genuinely require substantial explanation. This article is not short.
But the amount of content should follow the job the page needs to do.
Google says there is no ideal page length for generative search and recommends organizing content in ways that help readers, using clear sections and headings. Microsoft’s current AI-search guidance likewise points to clarity, useful structure, evidence, and completeness as qualities that can help content get referenced in AI-generated answers.
That should sound familiar because none of it is uniquely an AI tactic.
It is simply good information design.
Do this instead: Give the reader the useful answer early. Then provide context, evidence, exceptions, examples, and depth for the people who need it.
Authority sources:
- Google Search Central: Optimizing Your Website for Generative AI Features
- Bing Webmaster Blog: AI Performance in Bing Webmaster Tools
8. Please Stop Treating Internal Linking Like Clerical Work
[Insert the completed Evergreen Lantern hard-coded-domain meme here]
Evergreen Lantern says:
Please stop hard-coding the entire production domain into every internal link. We have twelve rooftops and somebody eventually has to migrate this shit.
This was the meme that kicked the larger idea into gear.
It came out of an internal-linking standards guide we use in dealership content work. The operational rule in that environment is to use relative paths for normal internal links because content may move across platforms, domains, and dealership rooftops.
That specific relative-URL requirement is an implementation rule for that environment, not a universal Google ranking requirement.
The larger SEO principle is universal.
Internal links help search engines discover pages, understand relationships, and interpret context. They also help humans move through the site.
Google recommends crawlable <a href> links and descriptive anchor text. Our internal dealership guide takes the practical implementation further: link contextually, avoid generic anchors, do not repeat the same exact-match text everywhere, link to the correct destination, avoid over-linking, and connect detail pages back to their hubs.
It also specifically recommends standard crawlable HTML links rather than relying on JavaScript click events for navigation.
Internal linking is architecture.
Treating it like an afterthought is how you end up with orphaned content, useless anchors, migration headaches, and important pages several logical universes away from their parent topics.
Do this instead: Build links for users first, make them crawlable, use meaningful anchors, connect related pages intentionally, and understand the URL conventions of the site you are actually working on.
Authority sources:
- Google Search Central: Link Best Practices
- Hendrick Automotive Group Internal Linking Best Practices, internal SEO and content reference guide.
9. Please Stop Using Canonicals Like a Magic Garbage-Disposal Button
[Insert Evergreen Lantern Meme #9]
Evergreen Lantern says:
Please stop pointing the canonical somewhere else every time you find a duplicate URL. Your canonical is a signal, not a fucking restraining order.
Canonicalization sounds simple until you actually have to debug it.
You have several duplicate or very similar URLs. You identify the preferred representative URL. You provide canonicalization signals.
Done, right?
Not quite.
Google describes canonical methods as ways to indicate your preference and explains that multiple canonicalization signals can reinforce one another. Google still determines canonical URLs within duplicate sets.
That is why canonical tags should not become a substitute for understanding why duplicate URLs exist.
If internal links, redirects, sitemap entries, parameters, canonicals, and page behavior all tell different stories, adding another canonical tag does not turn the situation into good architecture.
Do this instead: Decide which URL should exist and be surfaced, then make your signals coherent. Use redirects where redirects belong, canonicals where canonicals belong, clean internal links, and consistent sitemap URLs.
Authority source: Google Search Central: How to Specify a Canonical URL
10. Please Stop Chasing GEO Tricks Before Fixing Crawlability and Indexation
[Insert Evergreen Lantern Meme #10]
Evergreen Lantern says:
Please stop asking whether we need an
llms.txtfile before fixing therobots.txtfile you broke.
If Evergreen Lantern has a mission statement, this might be it.
There is nothing wrong with learning Generative Engine Optimization or Answer Engine Optimization.
There is something wrong with obsessing over speculative AI-search rituals while the site fails basic discovery and indexing requirements.
Google’s current guidance says pages need to be indexed and eligible for Search before they can appear in its generative AI features. It also emphasizes crawlability as part of generative-search visibility.
And Google is now remarkably explicit about several alleged AI-search hacks.
For Google Search, you do not need:
- an
llms.txtfile; - special AI markup;
- arbitrary “chunking” of content;
- rewritten machine-oriented prose;
- special schema designed solely for generative search.
Google says llms.txt currently neither helps nor hurts visibility in Google Search because Google Search ignores it.
That does not mean no other system will ever use it. It means you should understand the actual platform requirement before treating a new file as mandatory SEO infrastructure.
Do this instead: Confirm that important pages can be crawled, indexed, rendered, linked, understood, and served properly. Then experiment with emerging AI-search practices where evidence supports them.
Authority sources:
- Google Search Central: Optimizing Your Website for Generative AI Features
- Google Search Central: Introduction to robots.txt
11. Please Stop Buying Backlinks Because the Spreadsheet Says DR70
[Insert Evergreen Lantern Meme #11]
Evergreen Lantern says:
Please stop sending me a spreadsheet of 200 “high-DR” domains without checking whether a single one is remotely relevant.
Link building might be the SEO discipline most capable of turning a useful metric into a cargo cult.
Authority metrics can help with prospecting.
They can also make spectacular-looking spreadsheets full of websites nobody should actually want links from.
Google’s link-spam policy still prohibits buying or selling links for ranking purposes, excessive link exchanges, automated link creation, and other manipulative practices designed primarily to influence rankings.
Again, the basic lesson survived.
A link should make sense before its metric does.
Would a real reader plausibly encounter this link? Is the source relevant? Is the mention editorially defensible? Does the page itself have value? Would we still want the relationship if the third-party authority score disappeared tomorrow?
Those questions are less exciting than ordering 500 links.
They are also harder to automate.
Do this instead: Evaluate relevance, editorial context, audience, legitimacy, and actual usefulness before celebrating a metric. Build things worth referencing and relationships worth having.
Authority source: Google Search Central: Spam Policies for Google Web Search
Bonus #12: Please Stop Telling Me the Crawler Says It Is Fine. Click the Fucking Things.
[Insert Evergreen Lantern Bonus Meme #12]
Evergreen Lantern says:
Please stop telling me Screaming Frog says all the links are fine. Click the fucking things.
Automation is wonderful right up until it convinces us that nobody needs to look at the website.
A crawler can tell you an enormous amount.
It can identify status codes, missing tags, redirect chains, broken URLs, orphan candidates, duplicate elements, canonical problems, indexability signals, and thousands of other useful clues faster than a human ever could.
But the site still exists for humans.
Our dealership internal-linking guide explicitly treats automated link checking as a supplement rather than a substitute for manually clicking through important pages before and after publication.
That advice extends beyond links.
Open the page.
Use the navigation.
Click the CTA.
Read the copy on a phone.
Check the destination.
See whether the filter actually loads.
Look at what Google renders.
Sometimes ten seconds of behaving like a user reveals the problem your audit spreadsheet did not.
Do this instead: Automate the audit. Keep the human QA.
Authority source: Google Search Central: Link Best Practices, plus the Hendrick internal-linking QA standard cited above.
So What Does Any of This Have to Do With AI Search?
Quite a lot, actually.
Google now describes query fan-out, generative answers, grounding, AI Overviews, AI Mode, and generative visibility as part of the evolving Search experience.
Microsoft now reports when URLs are cited in AI-generated answers through Bing Webmaster Tools.
These are real changes.
They deserve real attention.
But Google’s own current guidance on generative search says the foundations remain the foundations.
Pages still need to be accessible.
Information still needs to be useful.
Site architecture still needs to make sense.
Content still needs a reason to exist.
Links still need to work.
Spam is still spam.
And now that machines can produce ordinary information in enormous quantities, being clear, useful, original, technically accessible, and actually worth referencing may matter even more.
The future of SEO is worth studying.
Just do not study it so hard that you forget SEO.
The Evergreen Lantern Gallery
Once all twelve graphics are finished, I would place a compact visual gallery here near the bottom of the article.
The article sections give each meme its context and rationale. The gallery gives readers an easy way to revisit, share, or save the set.
Suggested gallery order:
- Keyword stuffing
- Title-tag twitching
- Keyword-variation URLs
- Fake topical authority
- AI content sludge
- Optimization-score worship
- Burying the answer
- Internal-link chaos
- Canonical abuse
- GEO before crawlability
- Backlink-metric worship
- Human QA
If we later turn Evergreen Lantern into a recurring RBO series, this gallery can also become the first collection page.
FAQ: Do SEO Fundamentals Still Matter With AI Search?
Is traditional SEO still important for AI Overviews?
Yes. Google says its generative AI features are rooted in its existing Search ranking and quality systems. Crawlability, indexing, useful content, technical structure, and other established SEO practices remain relevant.
Do I need GEO in addition to SEO?
GEO and AEO can be useful labels for thinking about generative and answer-based search experiences, but Google currently treats optimization for its generative Search experiences as part of SEO rather than a completely separate discipline.
Does Google use llms.txt?
As of this writing, Google says Google Search does not use llms.txt and that having one neither improves nor harms visibility in Google Search. Other systems may make different choices, so evaluate it according to the platform you are trying to support.
Is AI-generated content bad for SEO?
Not automatically. Google focuses on the purpose, originality, quality, and value of the content rather than simply whether AI was involved. Using AI to create large amounts of low-value material intended primarily to manipulate rankings can fall under its scaled content abuse policy.
Final Action Item
Pick one page you care about.
Before worrying about GEO, citations, query fan-out, or the next AI-search acronym, check five boring things:
- Can search engines crawl it?
- Is it actually indexed?
- Does it satisfy one clear intent?
- Is there anything on it worth finding or citing?
- Can users and crawlers reach it through sensible internal links?
Then worry about the fancy stuff.
System > Hustle. Fundamentals > shiny objects.